摘要
目前有许多处理正面视觉人脸的识别方法,当有充分数量的有代表性的样本时,能取得较好的识别效果。然而当处理单样本识别问题时,现有的许多方法识别率将明显下降或甚至不适用。本文提出一种新的基于奇异值分解的训练图像增强的单样本人脸识别方法。为了从单训练样本中获取更多的信息,训练样本与其受扰动的少数较大的奇异值的重构图组合成新样本。然后进行Fourier变换,将Fourier频谱作为人脸识别特征,ORL人脸库上的实验结果表明了该方法的有效性。
At present there are many methods that could deal well with frontal view face recognition when there is sufficient number of representative training samples. However, few of them can work well when only one training sample per class is available. This paper proposes a new training sample enhancement method based on singular value decomposition to improve the performance of face recognition with a single training sample. In order to enhance the classification information of the single training sample, each training sample is combined with its reconstructed image gotten by perturbing a few significant singular values into a new version of the original sample. By using Fourier transform, the Fourier spectrum is used as feature for recognition. Experimental resuhs on ORL show the effectiveness of the method.
出处
《微计算机信息》
北大核心
2006年第10X期266-268,共3页
Control & Automation
基金
广东省自然科学基金资助(编号:05006593)